Publication types
Working paper — a substantial argument or model that is public before it is final.
Research note — a shorter position, question or architectural observation.
Experiment report — a bounded test with method, observations, outcome and limitations.
Evidence record — a machine- or human-readable record of what was observed in a specific test.
Status labels
Public draft means open for inspection and revision. It does not mean peer reviewed or independently replicated.
Published record marks a completed bounded observation for which supporting evidence has been retained or linked. The interpretation can still be corrected or superseded; the label is not a certification.
Active means the experiment or programme is still being developed or repeated.
Superseded means a later publication replaces the interpretation, method or design.
Evidence before interpretation
Where possible, we distinguish what the system observed from the conclusion drawn from it. A successful test supports a bounded statement about that test. It should not silently become a universal claim about a service, vendor or architecture.
Negative results count
Failures, refusals and inconclusive tests can be useful evidence. We do not treat a failed experiment as something to hide simply because it makes a cleaner story less convenient.
Limits belong on the page
Research pages should state important limits: what was not tested, what remains subjective, where evidence may age, and where independent replication has not happened.
Buyer policy is not our verdict
Atinamos can publish evidence about observed service behaviour. A buyer may use that evidence within its own policy. We do not turn payment for a test into a paid approval, certification or universal trust label.
Revisions and corrections
Substantive public research carries a version and publication date. Material corrections or changes should produce a new version and a visible revision note rather than silently rewriting the history.
Private work stays private by default
Labs is a curated publication surface, not a mirror of internal repositories. Operational details, unpublished test techniques, credentials and material that would make tests easier to game are not published merely because they exist internally.
Use of AI tools
Atinamos uses AI tools during software development, drafting and analysis. They are tools, not authors of record. Published claims remain the responsibility of Atinamos and should be judged against the evidence and limitations presented.